Learning Objectives

• Save and load trained model
• Create prediction API with Flask
• Deploy model to cloud

python
Run Inline
# Save model
import joblib
joblib.dump(best_model, 'models/model.pkl')
joblib.dump(scaler, 'models/scaler.pkl')
joblib.dump(le, 'models/label_encoder.pkl')
# Prediction API (app.py)
from flask import Flask, request, jsonify
import joblib
import pandas as pd
import numpy as np
app = Flask(__name__)
model = joblib.load('models/model.pkl')
scaler = joblib.load('models/scaler.pkl')
@app.route('/predict', methods=['POST'])
def predict():
    data = request.get_json()
    df = pd.DataFrame([data])
    df_scaled = scaler.transform(df)
    prediction = model.predict(df_scaled)
    probability = model.predict_proba(df_scaled)[0]
    return jsonify({
        'prediction': int(prediction[0]),
        'probability': float(probability[1]),
        'class': 'Positive' if prediction[0] == 1 else 'Negative'
    })
if __name__ == '__main__':
    app.run(host='0.0.0.0', port=5000)

Q1: What is serialization and why needed?

Saving trained model state (parameters, structure) to file. Enables reloading without retraining. joblib (sklearn), pickle, ONNX, PMML. Q2: How to handle model versioning?

Use MLflow or DVC. Tag models with version (v1.0), track parameters, metrics, and training data. Can rollback to previous versions. Q3: What is A/B testing for models?

Serve old and new model simultaneously. Compare metrics (conversion, accuracy). Gradually shift traffic to better model. Validates performance in production. Q4: How to monitor model in production?

Track prediction distribution, feature drift, label drift, response time, error rate. Set alerts for anomalies. Retrain when performance degrades. Q5: Deployment options?

Flask API on Heroku/AWS/GCP, FastAPI, Streamlit (demo), MLflow serving, Docker container, serverless (AWS Lambda). Q6: Docker deployment:

dockerfile
FROM python:3.9-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install -r requirements.txt
COPY . .
EXPOSE 5000
CMD ["python", "app.py"]

Q7: Test the API:

bash
curl -X POST http://localhost:5000/predict \
  -H "Content-Type: application/json" \
  -d '{"feature1": 5.1, "feature2": 3.5, "feature3": 1.4}'